{"schemaVersion":"1.0","type":"BacktestEvidence","id":"https://formiq.jp/evidence/bollinger-bandwidth-settings/data.json","evidencePage":"https://formiq.jp/evidence/bollinger-bandwidth-settings","language":"en","article":{"title":"Does the Bollinger BandWidth Squeeze Work? 4,032 Settings Tested","url":"https://formiq.jp/blog/bollinger-bandwidth-settings","published":"2026-08-28","updated":"2026-08-30"},"author":{"name":"Yuichi Saiki","alternateName":"斉木勇一","url":"https://formiq.jp/about/yuichi-saiki"},"verified":"2026-08-30","question":"Can the published 125-bar low or a fixed level define a Bollinger BandWidth squeeze?","finding":"Neither transferred. Defining a squeeze as the lowest value of the last N bars, the counts profitable in both 2024 and 2025 on the hourly chart were 37 of 64 at 20 bars, 31 of 64 at 50, 16 of 64 at 125 and 22 of 64 at 250, so the widely published 125-bar convention did not finish ahead. The strongest marginal was period 20, with 42 of its 64 combinations profitable in both years, yet the default-like setup — period 20, deviation 2, a 125-bar low and a 75-bar wait — took 58 trades and lost 70.6 pips in 2025. A fixed level does not travel between timeframes either: with period 20 and deviation 2 the 2025 median BBW was 0.2943 on M15, 0.6439 on H1 and 1.3693 on H4, so a level of 0.4 sits above the M15 median but below the H4 tenth percentile of 0.7642. BandWidth also scales with the deviation: moving from 1 to 2 doubled every measured value with a maximum error of 4.1e-14.","methodology":"Measured with Formiq's backtester. The only exit is the opposite signal, with entry and exit sharing one rule set. BandWidth uses closes, an SMA and the population standard deviation. No stop or target in the main grid.","scope":["USDJPY, 2025-01-01 to 2025-12-31, with 2024 run identically for comparison","15-minute, hourly and four-hour bars","Squeeze grid: 4 periods x 4 deviations x 4 contraction lookbacks x 4 waits = 256; fixed-level control: 4 periods x 4 deviations x 5 levels = 80. 336 settings x 3 timeframes x 4 windows = 4,032 runs","0.3 pip spread fixed, 0.1 lots, filled at the close"],"keyResults":[{"label":"Profitable in both years by contraction lookback, hourly","value":"20 bars 37/64, 50 bars 31/64, 125 bars 16/64, 250 bars 22/64"},{"label":"The default-like setup","value":"58 trades and −70.6 pips in 2025"},{"label":"Median BBW in 2025 at period 20, deviation 2","value":"0.2943 on M15, 0.6439 on H1, 1.3693 on H4"}],"limitations":["USDJPY in 2024 and 2025 only. On other pairs, equities or crypto both the distribution of BandWidth and the move after a break would change.","A squeeze here is the lowest value of the last N bars. A tenth-percentile definition, a run of low days, or a comparison against Keltner channels is a different condition.","Periods, deviations, lookbacks and waits are limited to the values listed; this is not a continuous optimisation over neighbouring numbers.","BandWidth is computed from closes, an SMA and the population standard deviation. Other applied prices or averaging methods were not measured.","Some four-hour settings trade only 3 to 10 times a year, which widens the uncertainty.","Fills use the close and the spread is fixed at 0.3 pips. Slippage, session-dependent costs and carry are excluded."]}